lation exposed is large and the exposures are great that it should be feasible to conduct environmental epidemiologic studies. The difficulty has been primarily with the lack of resources to conduct adequate exposure characterization or to collect sufficient outcome measures. It should be recognized, however, that for most sites the populations are too small to provide studies of sufficient power for risks to be detectable unless it is possible to combine data from a number of sites. Combining studies of small populations into meta-analyses might generate sufficient statistical power to reach conclusions, provided that the basic measures involved are comparable and that sound methods are used in all separate studies.
Several chapters explain that NCPs could be an important source of hazardous exposure. Some important pollutants are not regulated under a variety of acts, including the Safe Drinking Water Act, the Clean Air Act, and the Toxic Substances Control Act. These NCPs need to be identified and placed under appropriate regulatory control. Some preliminary toxicologic studies suggest that NCPs and so-called inert pesticide ingredients have important biologic properties of environmental persistence and mobility. More studies are needed to characterize the mixture of materials deposited as hazardous wastes and to give better estimates of their transport and fate in the environment. In the broadest sense, these unidentified, unregulated substances present a risk of unknown magnitude. The absence of evidence of their effects reflects the failure to conduct research; it should not be misconstrued as demonstrating that NCPs and “inert” pesticide components are without risk.
Where the potential for human exposure exists, exposure assessment should be conducted for important pollutants that can migrate from hazardous-waste sites. For the purposes of epidemiologic study, better characterizations of exposure are required than those usually available from engineering and hydrogeologic models and other estimates. Such models based on cursory local data often have been overemphasized while actual measurements or estimates of human exposure have gone undone. Models are only as valid as the assumptions and data on which they are based. Modeling needs to be improved both for health risks and for site assessment.
Many studies have focused on site-specific characterizations, even though pollutants do not respect boundaries. These site-specific investigations have often not proceeded to the steps of defining the populations at risk and quantitatively evaluating exposure to toxic contaminants. The characterizations of the sites more often reflect requirements of environmental engineering and site remediation than assessment of public health considerations. Whether the toxic con-